The effects of share repurchases on investment behavior

(1)

1

The effects of share repurchases on investment behavior

University of Amsterdam

Master Thesis, Academic year 2017-2018

Faculty of Economics and Business

Corporate Finance

Tom van der Meij, 10553355

Dr. Arnout Boot

Date 29-06-2018

Words: 12398

(2)

Statement of Originality

This document is written by Student Tom Petrus Eduardo van der Meij who declares to take full responsibility for the contents of this document.

I declare that the text and the work presented in this document are original and that no sources other than those mentioned in the text and its references have been used in creating it.

The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.

(3)

Abstract

This paper examines whether share repurchases have a negative effect on investments within a time horizon of four years. This research analyzes 60,000 share repurchases over a period ranging from 2000 to 2018, finding that repurchases have an overall negative effect on capital expenditure. At a sector-specific level, the effect of share repurchases is mainly negative on capital expenditure and negative on employment. Furthermore, there was no significant effect from using actual share repurchase periods instead of individual share repurchases. Finally, using accretive share repurchases in the regression, instead of negative pre-repurchase earnings per share (EPS) surprises, produced significant results, without affecting the impact of the negative pre-repurchase EPS surprise.

(4)

1 Introduction

Share repurchases have become an increasingly popular way to signal higher future performance to shareholders. This first started in the 1980s in the U.S. and has become popular in the EU since the 2000s. In recent years, its popularity has further increased. Between 2003 and 2012, companies in the U.S. used 54% of their earnings to repurchase their own shares, compared to 34% of earnings used for dividends (Lazonic 2014). Over the years there have been many papers on this subject. Most focus on the shareholder side and examine the positive effects on abnormal buy-and-hold returns, such as Ikenberry et al. (1995, 2000) and Chan et al. (2004). They have stated that share repurchase announcements are undervalued by the markets, which leads to an abnormal long-run shareholder returns. Furthermore, repurchases are popular among shareholders because companies tend to pay a higher price per share to comply with anti-manipulation and fraud provisions, such as the U.S. Securities and Exchange Act of 1934, to avoid lawsuits. However, share repurchases also have a negative effect. In his paper, Lazonic has suggested that the popularity of share repurchases is one of the causes that employment and overall economic prosperity has not grown after the financial crisis, even though the profitability of firms has increased. He argues that firms repurchase their shares, rather than investing their profits in growth opportunities. This presumption produced the following research question: Do share repurchases lead to a decrease in investments?

The number of papers that investigate these effects is limited, some of them include Grullon and Michaely (2004), Lie (2005), and Almeida et al. (2016). Grullon, Michaely and Lie found that repurchasing firms decrease their investments. Furthermore, their papers found no evidence of an increase in profitability for firms after repurchase. According to them, the evidence indicates that profitability decreases after repurchase. These papers give insight into how repurchases effect investment and how to measure these effects; however, their database predates the year 2000. There have been two large financial crises since then, which have led to new regulations and different managerial and investment behavior. Combined with the fact that share repurchases have increased in popularity since 2000, it can be argued that their data is outdated. Therefore, new research is needed to confirm their findings.

Almeida et al. (2016) wrote a more recent paper and used a dataset of quarterly data between 1988 and 2010. They focused on how repurchases have been used as a tool to influence earnings per share (EPS) to meet or beat targets. The gap between pre-repurchased EPS and forecasted EPS was used as a predictor for accretive share repurchase. They found that firms use these accretive share repurchases to increase EPS so as to meet EPS targets that would have been missed if they had not performed the repurchase. The paper showed that these accretive share repurchases are associated

(6)

with reductions in employment, investment, and cash holdings. This conclusion was reached by comparing the average investment outcome from one year before to one year after repurchase. Furthermore, they have suggested that managers are willing to exchange employment and investments for accretive repurchases, to meet their EPS target.

The Almeida et al. (2016) paper forms the basis for this research. Its data are relatively young compared to other research, and it uses an identification strategy that is less likely to be plagued by endogeneity problems. However, their paper has some limitations. First, they only look at the effect of repurchases by comparing the year before and year after repurchase, not allowing for potential reinvestment in the following years. Furthermore, Ikenberry, Lakonishok, and Vermaelen (1995, 2000) and Chen and Wang (2012) asserted that the effects of repurchase last for four years; thus, using a horizon of four years instead of one should be more accurate. Furthermore, previous research has used either share repurchase announcements or actual repurchases. However, Lie (2015) has shown that the effects of repurchase occur when using actual repurchases. Yet, papers that use actual repurchases look at repurchase as individual events, although repurchases are performed in succession. One might argue that using actual repurchase periods should lead to better results. Therefore, this paper also looks at the different outcomes when comparing individual repurchases to share repurchase periods. Finally, this research looks at the direct effect of accretive share repurchase, when compared with the negative pre-repurchase EPS surprise used by Almeida et al. and Hribar et al. (2006). By doing so, this paper offers a unique view on the negative effect on investments that potentially occurs after share repurchase.

The research presented here analyzed 60,000 share repurchases, dating from 2000 until 2018. Capital expenditures, employment, and research and development (R&D) expenses were used to measure the outcome variables of investment. To test the effect of repurchase, the average investment outcomes from the four years before and after the repurchases were used. The results show that repurchase has a significant effect on changes in capital expenditure and employment, when comparing four years before with four years after repurchase. These effects are generally negative, but vary when examined at the sector level and by credit-strength group. The results of this research are not completely aligned with those of the previous literature, since they demonstrate that repurchases have an overall negative effect on capital expenditures, employment, and R&D. This can be explained by the fact that this research uses a horizon of four years, which allowed firms to compensate for underinvestment in one year by overinvesting the following year. This research shows that different sectors affect firms in a way that is not captured in the control variable. The results also indicate no significant effect from using actual share repurchase periods instead of individual share repurchases. Finally, the use of a term for accretive share repurchases produced significant results, without impacting the effect of the negative pre-repurchase EPS surprise.

(7)

This paper is divided as follows. In Chapter 2, the extant literature and hypotheses are discussed. Chapter 3 elaborates the methodology of this research, and Chapter 4 describes the data that are used for this study. In Chapter 5, the regression results of this paper are analyzed, and Chapter 6 discusses the robustness test. Chapter 7 provides a summary of this paper and presents a conclusion. This chapter also discusses biases in the data, missing data, and other errors and surprises, as well as recommendations for future research.

(8)

2 Literature review

Most papers that research repurchase have focused on its effect on abnormal buy-and-hold returns. Although these studies differ slightly from this paper, the setup of these research projects formed the foundation of this research. A few papers have examined the effect of repurchases on investments; they are used, as well. In this chapter, the results of the most relevant studies are discussed, and a hypothesis is formed.

Share repurchases have increased in popularity in recent years. The effect of share repurchase has been examined by many researchers, such as Ikenberry, Lakonishok, and Vermaelen (1995, 2000) and Chen and Wang (2012). These studies have mainly focused on the positive effects of share repurchase. However, repurchasing shares also has negative effects. Lazonick (2014) examined these negative effects; he observed a lack of growth in employment and of overall economic prosperity, even though corporate profitability increased. He argued that this is caused by the decision of companies to repurchase their shares instead of investing their profits in growth opportunities. Lazonick examined 449 firms in the S&P 500 that were publicly listed from 2003 through 2012. He found that these companies used 54% of their earnings to repurchase their own shares and another 37% for dividends. Lazonick suggested that this is caused by stock-based instruments that make up the majority of executives’ pay. This causes an incentive to repurchase, which leads to a short-term share-price increase and, thus, higher executive pay. His paper shows how high repurchase expenses have become and how they could damage the economy of the United States of America He urged government and business leaders to take action, to avoid further economic damage. The number of papers researching these negative effects is limited and those that do examine the relation between share repurchases and investments are relatively outdated. Combined with the increase in popularity of share repurchase, Lazonicks findings give an incentive to do more research with recent data.

2.1 Reasons for repurchase

It is important to look at the shareholder side of the repurchases to understand why they are not opposed even though the trend of share repurchase can be harmful. This can be explained partly by the positive relation between share repurchases and future returns that several papers have found. The basic theory behind this is based on the information imbalance between shareholders and firms. Firms can better judge when their shares are undervalued, which allows them to repurchase their shares when they are relatively low in price. This is seen by the market as a positive signal for future returns, which leads to an increase in the share price and, thus, an increase in shareholder returns. Shareholders do not oppose share repurchases because they see their returns increase.

(9)

Theo Vermaelen has written several papers about this increase in shareholder returns through share repurchase. One of his earlier works is ‘Common share repurchases and market signaling: An empirical study’ (1981). For this research, Vermealen used 131 tender offers made by 111 U.S. firms from 1962-1977 and 243 open market purchases made by 198 U.S. firms from 1970-1978. In this paper he stated that firms offer a higher price per share during repurchases to comply with the anti-fraud and anti-manipulation provisions of the Securities Exchange Act of 1934, to avoid lawsuits. Furthermore, he found that the results of an open-market purchase are less conclusive but still consistent.

Later he wrote two papers with Ikenberry and Lakonishok (1995, 2000), in which they research the effect of repurchases on the average abnormal four-year buy-and-hold returns, measured after the initial announcement. The first study examined U.S. data from 1980-1990 and the second paper focused on Canadian data from the 1990s. The researchers in both cases claimed that the market undervalues the effect of repurchases and that repurchases lead to an increase in abnormal returns. Vermealen is currently writing a paper with Manconi and Peyer (2018), in which they are researching a sample of 9,000 repurchases announcements from 1998 to 2010, from 31 countries outside the U.S. They have found similar results to Vermealen's U.S.-based studies; repurchases are associated on average with significant positive short-term and long-term excess returns. They state that excess returns depend on the likelihood of undervaluation, efficiency, and liquidity of equity markets, as well as the popularity of stock-option compensation. However, they have not found that long-term excess returns are a compensation for takeover risk or that they have become less significant in recent years, in contrast with the U.S. markets findings. These papers help to explain how share repurchases lead to an increase in shareholder returns and provide insight into how to observe and analyze share repurchases.

2.2 Effect on investments

As mentioned above, most literature focuses on the positive relation between share repurchases and abnormal returns, but they do not examine the potential negative effect on investments, as mentioned by Lazonick (2014). The basic theory behind this negative effect is that firms use cash that could be used for investments to repurchase shares, with the goal of giving a positive signal to the market. When firms decrease their investments too much, they are underinvesting. Underinvesting can lead to a loss in firm value, if it affects the asset side of the firm's balance sheet. However, if the cash loss is compensated with new debt, and the firm does not underinvest, then the repurchase only changes the balance sheet's ratios and should not have a negative effect on firm value. A few papers have been written that look more closely at the negative effects of share repurchases on investments.

(10)

One of these studies is the 2004 paper of Grullon and Michaely. Their goal was to gain a better understanding of the economic motivations behind firms’ decisions to repurchase shares. They examined two major motivations for share repurchases: namely, to signal better prospects and to reduce the amount of free cash flow at the managers disposal. This study used a sample of 4,443 open-market share repurchase announcements for the period between 1980 and 1997 from U.S. firms. Grullon and Michaely (2004) found decreases in investment caused by share repurchase. They also found that operating performance does not increase after an announcement of open-market share repurchase programs. However, Grullon and Michaely (2004) did find that firms that perform repurchases experience a significant reduction in systematic risk and cost of capital in comparison with non-repurchasing firms. They also found that the market reacted more positively to share repurchase announcements from firms that were more likely to overinvest. Furthermore, the researchers noted that investors tended to underreact to repurchase announcements because they underestimated the decline that would occur in the cost of capital. That research provided insight into how to measure share repurchases and the effect on investment, which can be used for this research. Grullon and Michaely (2004) also offered an understanding of how repurchases lead to reduction in systematic risk and cost of capital, which indicates that firms attract cash more easily after repurchase. This is important to consider when researching the effect of repurchase after more than one year. It was also stated that the market reacts more positively to share repurchases when firms are likely to overinvest. This could suggest that the market understands the negative effects of share repurchases on investments and the reaction is more positive when a firm is not likely to be underinvesting. However, the paper uses data from 1980 to 1997. The problem with this is that there have been two large financial crises since then, which have led to new regulations and different managerial and investment behavior. There has also been an increase in the popularity of share repurchase since 2000. These reasons would indicate that the research described above is outdated; therefore, more recent studies are needed.

Lie (2005) followed Grullon and Michaely (2004) by reexamining operating performance after 4,729 open-market share repurchase program announcements in the period from 1981 to 2000. For this research, Lie used the same SDC Mergers and Acquisitions database as Grullon and Michaely (2004). He differed by using quarterly instead of annual data and by making a distinction between repurchase announcements and actual repurchases. Lie found that operating performance improves and there is a favorable reaction from the capital market to the earnings announcements, after the program announcements. Furthermore, he stated that these improvements in operating performance and the positive reaction to earnings announcements only occur for firms that actually repurchase their shares in the same quarter. This suggests that the actual repurchase, rather than the

(11)

announcement, leads to future performance improvements. Lie, as Grullon and Michaely did, offered insight into how to measure the share repurchases and the effect on investment. Furthermore, the paper described how share repurchases can best be measured. However, the data set of Lie dates from before 2000, so it is likely outdated and needs new research.

2.3 Identification strategy

To identify the relationship between share repurchase and investment, it is important to know whether share repurchases could also be used as a tool for misleading shareholders. If an underperforming firm could use share repurchase to send a positive signal, with the goal of gaining additional access to external finances, the effects on investment would differ. However, Chan, Ikenberry, Lee, and Wang wrote a paper about the misleading use of share repurchases in 2010. In this paper, they examined whether mangers use share repurchases as a tool to mislead investors. They focused on examples in which managers seemed to be under heavy pressure to boost the share price and might have used repurchased announcements as a false signal. They found that these firms experienced no otherwise observed improvement in economic performance. Moreover, managers in such firms had a comparatively higher exposure to stock options and a great sensitivity to both future equity dilution and stock valuation. However, they found that the number of these managers is limited. Taking this into account, it is not necessary to distinguish between the effect of share repurchases on investment between firms under financial constraint and financially unconstrained firms.

A different approach towards share repurchase considers how it is used to meet and beat EPS targets. This method has been described by Hribar et al. (2006). They examined the conditions in which repurchases lead to an increase in EPS and they documented the frequency of accretive repurchases in the period from 1988 to 2001. By performing accretive repurchase, the outstanding shares decreased in such a way that the EPS increased significantly. They define a repurchase as accretive if it led to an increase in EPS of at least US $0.01. This increase was calculated by comparing the EPS and the adjusted EPS, which was the EPS if there had been no share repurchases. They found that firms who would have missed the forecasted EPS without repurchase, had a disproportionately large number of accretive share repurchases among themselves. For this research, Hribar et al. (2006) used a share repurchase sample of 13 years, from 1988 to 2001, which included only U.S. firms listed on the AMEX, NASDAQ, or NYSE exchanges. Hribar et al. (2006) helped to find an identification strategy for the effects of share repurchases that is not plagued by endogeneity problems and, thus, can be used in this paper. However, since their dataset is outdated, a new dataset is required.

Almeida, Fos, and Kronlund (2016) used these findings for their own research on the effects of share repurchases on investments. Their research was based on S&P Compustat, CRSP, and IBES

(12)

data, using quarterly data between 1988 and 2010. They used a pre-repurchase EPS surprise and an indicator for a negative pre-repurchase EPS surprise as an independent variable. The pre-repurchase EPS surprise is the difference between the adjusted EPS and the forecasted EPS. It was found that the probability of accretive share repurchases are much higher for firms that would have missed their EPS forecast without the repurchases, compared to firms that would narrowly beat their EPS forecast. They showed that these accretive share repurchases are associated with reductions in employment and investment and a decrease in cash holding. This conclusion was reached by comparing the average investment outcome one year before and one year after the repurchase. Their paper suggested that managers of these firms are willing to trade on employment and investments to repurchase shares in order to meet EPS forecasts. Almeida et all forms a foundation for this research. Its data are relatively young compared to other research, and it uses an identification strategy that is less likely to be plagued by endogeneity problems. However, their paper has some limitations. First, it looks at the average investments only one year before and one year after the share repurchase. The problem with this method is that it does not consider that underinvestment in one year can be compensated for by overinvestment in the next year. As mentioned by Grullon and Michaely, share repurchases can lead to easier access to external finance, which makes it possible to compensate for underinvestment. Several papers, such as Chen and Wang (2012) and Vermaelen Ikenberry and Lakonishok (1995;2000), have suggested that the effects of share repurchases last until four years after the repurchase. Therefore, it would be more informative to look a four-year horizon. Therefore, the model of Hribar et al. needed to be updated.

2.4 Hypotheses and conceptual framework

This research focuses on the investment effect of share repurchases. It continues the work of Lie (2005), Hribar et al. (2006), and Almeida et al. (2016). To measure investment, the variables of capital expenditures, employment, and R&D are used as outcome variables, as described by Almeida et al. To test the effect of share repurchases on investment, several hypotheses are formed.

The first hypothesis is the basis of this paper and is used to test whether share repurchases have a negative effect on investments. This hypothesis is based on Almeida et al. (2016). They presumed that firms spend too much cash on repurchases, which led to less cash for investments, which caused underinvestment. For the first hypothesis, a similar presumption is taken, but with a larger time horizon. To test this, the average of capital expenditures, employment, and R&D over a period of four years after the repurchase are compared to the average of four years before the share repurchases. The expectation is that the effect of share repurchases is significant and negative on investment.

(13)

The second hypothesis is based on the way repurchases are measured. Previous literature has used either announcements of repurchase programs or actual repurchases. According to Lie (2015), repurchases only lead to operating performance improvement and positive earnings announcement returns if the firms who announce the repurchases actually do repurchase these shares during the same fiscal quarter. This suggests that actual repurchases are a better measurement than repurchase announcements. However, those who use actual repurchases measure individual share repurchases and do not correct for the fact that most repurchases are done in periods of successive repurchases. Therefore, a hypothesis was formed that states that the use of share repurchasing periods produces more significant results than the use of individual repurchases.

H2: Using share repurchasing periods produces more significant results, compared to the use of

individual share repurchases.

The third hypothesis is based on the accretive share repurchase. According to Hribar, Jenkins, and Johnson (2006) accretive repurchases are performed when negative EPS surprises occur, which could be countered with the accretive repurchases. Both Hribar, Jenkins, and Johnson (2006) and Almeida et al. (2016) have used the EPS surprises as a predictor for accretive repurchases. However accretive share repurchases can also be used directly in the model. Adding this term to the regression should have a significant negative effect on investments. Furthermore, since the negative pre-repurchase EPS surprise is used as predictor for accretive repurchases, the significance of the surprise should reduce since the effect is already captured by the accretive indicator.

H3: Adding the accretive share repurchase term leads to a decrease in the effect of the EPS surprise

term.

2.5 Summary of changes

This research continues the work of Almeida et al. (2016), but differs in the following ways. First, a different time period is used; this paper uses data from 2000 until 2018, rather than 1988-2010. This period was chosen for contemporary relevance and because share repurchases have increased in popularity since the 2000s. Second, this paper compares average investment outcomes over the four years before repurchase with average investment outcomes in the four years following the repurchases. Almeida et al. used one year, which does not account for the ability to compensate underinvesting in one year by over investing in the next. Third, this research adds several control variables that measure whether the decrease of cash by repurchases is compensated for by issuing new debt or equity.

(14)

Investments Share repurchases Share repurchase periods Accretive share repurchases Figure 1

The common concepts of share repurchase and the three hypotheses.

H1 H2

H3

- -

(15)

3 Methodology

This chapter explains how the research question was answered. The regressions are further elaborated upon and regression formulas provided.

3.1 Univariate regression

This paper uses a panel data analysis with CRSP, Compustat, and IBES data from the period 2000 to 2018, to test whether share repurchases lead to a decrease in investments. Furthermore, the use of share repurchase periods and the direct use of accretive share repurchases is examined. A total of four regressions are used to answer the research question. The first regression is similar to the univariate OLS regressions of Almeida et al. (2016). This regression was used to see whether a broader time horizon affects the univariate regression. The results indicate the effect of the investment variables. The expectation was that repurchases have a negative effect on the investment variables. However, it was also expected that the effect is smaller than that mentioned by Almeida et al., because underinvestment can be compensated for in the following years. In the first regression, capital expenditures, employment, and R&D were used as outcome variables for investments. The outcome variables were measured in the difference between the average of four years after the repurchase and the average of the four years before the repurchase, normalized by the lagged assets. As in Rauh (2006), two control variables were added to the investment regression: Tobin’s Q and Cashflow divvied by assets. The regression was further corrected for year-quarter fixed effects. This resulted in the regression stated below.

Regression 1: 𝐼̅𝑖,(𝑡+1,𝑡+16)− 𝐼̅𝑖,(𝑡−16,𝑡−1)= α + β1𝑅𝑒𝑝𝑢𝑟𝑐ℎ𝑎𝑠𝑒𝑠𝑖𝑡+ β2𝑋 + 𝜃𝑡+ 𝜀𝑖𝑡 In this regression, time (t) is 0 at the moment of the share repurchases; I stands for investment outcomes. Repurchase is quarterly repurchases divided by total assets. X is the control variables: Tobin’s Q and cashflow divided by assets. θ is year-quarter fixed effect.

3.2 EPS identification strategy and control variables

The shortcoming of the first regression is that it is subject to endogeneity problems. One of these problems is omitted-variable bias. If the investment opportunities of a firm are not profitable, then they will use less cash for investment and increase repurchases. There could also be reverse causality; if the firm spends more on investments, then less cash remains for repurchases. These problems are also mentioned by Almeida et al. (2016). To take these problems into account, Almeida et al. introduced an identification strategy that uses the discontinuity in the level of share repurchases, based on the pre-repurchase EPS surprises discovered by Hribar et al. (2006). This paper also uses this EPS identification strategy by using the pre-repurchase EPS surprise and the negative pre-repurchase

(16)

EPS surprise.

Several control variables were used; these variables were added to control for omitted-variable bias. First are the control omitted-variables also used by Almeida et al. (2016): a dummy that indicates whether the firm payed dividends in the previous year, a coefficient for return on assets, a variable for quarterly stock returns, and a variable for the ratio of cash to assets. Two new control variables were added to control for the fact that firms can compensate for a loss in cash by repurchasing or by issuing new debt or equity. These two variables are a coefficient for increase in debt in the current quarter and a coefficient for shares that were issued in the current quarter. These two variables were expected to have a positive effect on investment, since an increase in debt and equity can compensate for a decrease in cash resulting from share repurchases, thus preventing underinvestment due to cash shortage.

3.3 Basis regression

The second regression is a combination of Almeida et al. (2016) and Hribar et al. (2006), plus the control variables previously mentioned. This regression forms the basis for the research presented in this paper. It tests how changes in investment outcomes are influenced by repurchases and the negative pre-repurchase EPS surprise. The regression equation is given below. A negative sign was expected for both repurchases and the negative pre-repurchase EPS surprise, consistent with Almeida et al.

Regression 2: 𝐼̅𝑖,(𝑡+1,𝑡+16)− 𝐼̅𝑖,(𝑡−16,𝑡−1)

= α + β₁𝑅𝑒𝑝𝑢𝑟𝑐ℎ𝑎𝑠𝑒𝑠_𝑖𝑡+ β₂𝑆𝑢𝑒_{𝑎𝑑𝑗,𝑖𝑡}

+ β₃𝑆𝑒𝑢_{𝑎𝑑𝑗,𝑖𝑡}𝐼_{𝑁𝑒𝑔𝑎𝑡𝑖𝑣𝑒 𝑆𝑢𝑒}_{𝑎𝑑𝑗,𝑖𝑡} + β₄𝑥 + 𝜃_𝑡+ 𝜀_𝑖𝑡

In this equation, Sueadj indicates the pre-repurchase EPS surprise, I Negative indicates that there was

a negative pre-repurchase EPS surprise, X stands for the different control variable (payment of dividends in the previous year, return on assets, debt issued, shares issued, quarterly stock returns, and the ratio of cash to assets) and the θ is year-quarter fixed effects.

3.4 Individual repurchases versus repurchase periods

The shortcoming of the previous regression is that it does not correct for multiple repurchases in succession. As shown in Table 2, share repurchases are done successively. To take this into account, the model is adjusted to describe the effects of a period of repurchase by comparing the four years before the beginning and the four years after the relevant share repurchase period. A maximum of 12

(17)

quarters were taken for the duration of the share repurchase. To correct for endogeneity, the same control variables are used as in regression 2.

Regression 3: 𝐼̅𝑖,(𝑡𝑝𝑒𝑟+1,𝑡𝑝𝑒𝑟+16)− 𝐼̅𝑖,(𝑡𝑝𝑒𝑟−16,𝑡𝑝𝑒𝑟−1)

= α + β₁𝑅𝑒𝑝𝑢𝑟𝑐ℎ𝑎𝑠𝑒𝑠_𝑖𝑡+ β₂𝑆𝑢𝑒_{𝑎𝑑𝑗,𝑖𝑡}

+ β3𝑆𝑒𝑢𝑎𝑑𝑗,𝑖𝑡𝐼𝑁𝑒𝑔𝑎𝑡𝑖𝑣𝑒 𝑆𝑢𝑒𝑎𝑑𝑗,𝑖𝑡 + β4𝑥 + 𝜃𝑡+ 𝜀𝑖𝑡

In this regression, time (t) is 0 during the share repurchase period.

3.5 Accretive repurchases

The previous model uses negative pre-repurchase EPS surprise as a predictor for accretive share repurchases, which is consistent with Almeida et al. (2016). According to Hribar, Jenkins, and Johnson (2006), accretive repurchases have a significant effect on investment because a large share of cash holdings is needed to pay for such a repurchase to increase EPS, leading to a decrease in investments. It can be argued that accretive share repurchases can also be used directly in the regression; this is tested in the fourth regression. The expectation was that the indicator for accretive repurchases has a similar value as the negative pre-repurchase surprise in the second regression. It is further expected that the negative pre-repurchase loses its significance, since it was a predictor of accretive repurchases. When adding this indicator, the following regression is created.

Regression 3: 𝐼̅𝑖,(𝑡+1,𝑡+16)− 𝐼̅𝑖,(𝑡−16,𝑡−1)

= α + β₁𝑅𝑒𝑝𝑢𝑟𝑐ℎ𝑎𝑠𝑒𝑠_,𝑖𝑡+ β₂𝐼_𝑎𝑐𝑐+ β₃𝑆𝑢𝑒_{𝑎𝑑𝑗,𝑖𝑡} + β4𝑆𝑒𝑢𝑎𝑑𝑗,𝑖𝑡𝐼𝑁𝑒𝑔𝑎𝑡𝑖𝑣𝑒 𝑆𝑢𝑒𝑎𝑑𝑗,𝑖𝑡 + β5𝑥 + 𝜃𝑡+ 𝜀𝑖𝑡

(18)

4 Data description

This research uses data from CRSP, Compustat and IBES, from the period of 2000 to 2018. This chapter describes the acquiring and adjustment of these data.

4.1 Sample selection

The dataset is panel data that consist of quarterly observations of both market and accounting values from U.S. firms. The dataset was obtained from Compustat, IBES, and CRSP. From Compustat, both annually and quarterly Capital IQ North American fundamentals were used. The data that were acquired from IBES were from Detail History - Detail File with Actuals. From the CRSP database, the files that contain the U.S. treasury inflation Indexes, as well as stock and security, were used. The sample period begins in 2000-01 and ends at 2018-01. Firms that are regulated utilities (SIC codes 4800-4829 and 4910-4949), financial firms (SIC 6000-6999), and firms with less than 16 observations were removed from the database. This is consistent with Almeida et al. (2016). Survivorship bias was prevented by observing both active and inactive firms. All data were winsorized with 5% to reduce the effect of large outliers. After these adjustments, the complete, merged database consisted of 537,447 observations.

4.2 Summary statistics

Panel A of Table 1 reports the summary statistics of the repurchase activities. In 11% of all firm quarters, firms repurchased their own shares (resulting in a net positive repurchase). When conditioning these net positive repurchases on firm-quarters, the average value in dollars of these repurchases was $87.81 million. This is equal to 1.15% of the total lagged shares outstanding and 1.61% of the total book assets at the end of the previous quarter. When compared to the data of Almeida et al. (2016), the number of shares repurchased are shown to have decreased. However, the average value in dollars of the repurchases and the ratio of repurchases to assets increased because of the relatively large values in the top percentiles. The numbers below the P99 are consistent with those of previous research. Panel B describes the statistics of the earnings surprises. According to these statistics, the earnings forecasts are accurate, on average, since the median is zero. Furthermore, the positive surprises (53.1% of the sample) are more likely reported, compared to the negative surprises (43.7% of the sample). The number of positive surprises is slightly higher than those of previous research but is otherwise consistent. Panel C reports the statistics on the firm characteristics in the sample. The results of these statistics are consistent with other papers that have used Compustat data. To gain insight into how these variables differ across sectors, the four largest sectors were isolated, to compare their summary statistics with Table 1. These sectors are mining, manufacturing,

(19)

TCEGS (Transportation, Communications, Electric, Gas and Sanitary), and service; together these are 467,312 of the 542,058 observations, or roughly 86%. Panel A shows that the number of repurchases relative to firm-quarters is roughly the same, with the exception of the mining sector, which had repurchases of only 5% of all firm-quarters. The average value of these repurchase is higher in the manufacturing and the TCEGS sectors, while it is lower in the mining and service sector. This can be explained by the fact that these sectors have more asset value, on average, even though their repurchase value divided by assets is relatively lower. The service sector has the highest average rate, at 1.99%, and mining the lowest, at 0.59%, when dividing repurchases by total assets.

According to Panel B, the sectors are less accurate in earnings forecasting, since their median is not zero. The number of positive surprises is lower, except for manufacturing, which is equal, and the service sector, which is higher.

Panel C shows a large variety of firm characteristics across sector. As mentioned above, the size of the average firm in each sector is widely different, as are their shares of cash, liabilities, and share of investments per assets. Capital expenditures are, on average, larger for the mining and the TCEGS sectors, while the service sector has a relatively larger share of employees.

4.3 Successive share repurchases

Table 2 shows the quantitative data for the share repurchases. The first panel describes the different length of repurchase periods divided across the observations. It shows that a repurchase period lasts, on average, between seven and eight quarters, but has a median of four quarters. This means that there are relatively large values in the top percentiles. The second panel describes the accretive share repurchases. This panel shows that 43.38% of all repurchases are accretive. The average number of successive accretive repurchases is lower than that normal repurchases. Furthermore, the maximal number of successive accretive repurchases is also lower than those of regular repurchases. This can be explained by the fact that accretive repurchases are relatively costly and, thus, harder to do in succession. To better view repurchase periods across sectors, Appendix II shows a sector-specific table. This demonstrates that the distribution of successive repurchases is similar across sectors. The distribution of accretive repurchases is also roughly similar across sectors, except for the mining sector, which has a much lower percentage of 18.40% of its total repurchases. The percentage of accretive repurchases divided by assets varies across sector, which can be explained by the difference in firm characteristics. The distribution of successive accretive repurchases is similar across sectors, except in the mining sector, for which successive accretive repurchases are much lower.

(20)

Table 1

The following observations are based on firm-quarters and defined as in Almeida et al. (2016). Panel A reports the summary statistic of share repurchase activities. The repurchase dummy ("Binary") has a value of 1, if the total quarterly share repurchases are not zero or missing. Repurchases in millions (Million Dollars) are the total quarterly share repurchases multiplied by average daily closing share price during the quarter. "Repurchased shares/Shares outstanding" is the total quarterly share repurchases divided by common shares outstanding. "Repurchases/Assets" is repurchases in millions divided by total assets. Panel B shows the summary statistics of the earnings surprises. An earnings surprise is calculated by subtracting the median of the EPS forecast at the end of the quarter from the reported EPS. This difference is divided by the average daily share price during the quarter. The indicators are dummies that show whether the surprise is negative, zero, or positive. Panel C reports summary statistics about firm characteristics. All measures that are scaled on asset use lagged assets from the end of the previous quarter. Return on assets was calculated by dividing net income by lagged assets. "Q" is measured by taking the book value of liabilities, adding the market value of common equity, and dividing this by the book value of assets [(atq – ceqq + marketcap)/atq]. Cash flow is calculated by adding depreciation to net income and dividing this by lagged assets. Market-to-book value is the market value of common equity divided by the book value of common equity [marketcap/(seqq-pstkq)]. Accrual is calculated by the following formula accrualst=(actt−actt−1)−(chet−chet−1)−(lctt−lctt−1)+(dlct−dlct−1)−dp,

as given by Gormley & Matsa (2016). Stock returns are the quarterly raw stock returns from CRSP. The dividend payer is an indicator that shows whether a firm paid any dividends in the last four quarters (this includes the current quarter.) The issuance of equity is the purchase of common and preferred stock minus the (net) repurchases. The issuance of debt is the change in total debt.

Panel A:Repurchase statistic mean SD P1 P5 P25 P50 P75 P95 P99 N Binary .11 .31 0 0 0 0 0 1 1 542,058 Million Dollars 87.81 243.91 0 .03 0.48 5.92 49.22 478 1664 59,999 Repurchase shares/Shares outstanding 1.15% 1.76% 0 0 0 0.5% 1.44% 4.47% 10.71% 60,346 Repurchases/Assets 1.61% 2.64% 0 0 0.8% 0.61% 1.92% 6.80% 15.77% 59,825

(21)

Panel B: EPS surprise mean SD P1 P5 P25 P50 P75 P95 P99 N Earnings surprise/ stock price -0.94% 5.80% -38.89% -7.95% -.82% 0% -.38% 3.38% 15.85% 159,919 Positive ES 53.13% 49.90% 0 0 0 1 1 1 1 174,959 Negative ES 43.72% 49.60% 0 0 0 0 1 1 1 174,959 Zero ES 3.15% 17.46% 0 0 0 0 0 0 1 174,959 Panel C: Control variable mean SD P1 P5 P25 P50 P75 P95 P99 N Market Cap ($M) 2,229 7,667 0.25 1.69 20.54 131 845. 10,600 57,974 458,892 Asset ($M) 2,442 9,394 0.006 0.501 19.06 133.7 852.3 10,816 49,658 498,608 Cash/Asset 24.20% 32.60% 0.00% 0.13% 2.93% 11.30% 33.40% 84.00% 198.% 468,075 Total liabilities/Asset 117.80% 386.50% 1.15% 5.79% 26.20% 49.00% 73.90% 264.% 3300.% 468,350 Capex/Asset 1.73% 2.89% 0.00% 0.00% 0.29% 0.76% 1.78% 7.01% 18.70% 446,020 R&D/Asset 0.% 0.02% -0.22% 0% 0% 0% 0% 0% 0% 432,811 Employees/Asset 1.89 3.55 0 0.03 0.37 0.87 1.86 3.95 26.79 394,472 ROA -14.00% 62.70% -513.30% -57.10% -5.49% 0.00% 1.67% 5.51% 21.60% 471,785 Q 6.60 25.50 0.40 0.69 1.11 1.61 2.91 16.30 219.80 454,310 Market-to-book 2.20 12.58 -72.86 -4.94 0.77 1.72 3.44 12.77 65.64 456,206 Cash flow/Asset -7.21% 41.60% -329.50% -41.50% -4.12% 1.28% 4.44% 11.80% 36.60% 421,962 Accruals -22.15 105.80 -759.80 -136.00 -8.33 -0.48 0.38 25.96 204.20 425,611 Crisis dummy 0.52 0.50 0 0 0 1 1 1 1 537,009 Stock returns 2.59% 29.10% -80.70% -45.00% -11.70% 1.43% 15.80% 51.30% 107.10% 284,338 Dividend payer 0.32 0.47 0 0 0 0 1 1 1 537,009 Equity issuance/Asset 0.49% 1.80% -0.19% 0% 0% 0% 0% 0.94% 12.50% 430,306 Debt issuance/Asset 5.94% 38.10% -73.90% -12.70% -2.01% 0.24% 3.54% 32.20% 300.0% 468,142

(22)

Table 2

The distribution of successive repurchases measured with a dummy variable: 1, if there is a repurchase in the current period; 2, if there was a repurchase in the previous period and in the current period; 3, if there was a repurchase in n-2 n-1 and n, and so forth. Accretive repurchase is indicated with a dummy if the repurchase led to an EPS increase of at least US $0.01. When dividing accretive repurchases with assets, the lagged asset of the previous quarter is used. For the distribution of successive accretive repurchases, the same method is used as for the first variable with accretive repurchases only.

Repurchase quantities mean SD P1 P5 P25 P50 P75 P95 P99 N

Distribution of successive repurchases 7.57 8.55 1 1 2 4 10 30 ≥32 60,043 Distribution of accretive repurchases 43.38% 49.56% 0 0 0 0 1 1 1 41,003 Accretive repurchases/Asset 3.15% 3.20% .21% .42% 1.07% 2.05% 3.94% 7.11% 15.88% 17,767 Distribution of successive accretive repurchases 4.53 5.49 1 1 1 2 5 17 28 17,788

(23)

5 Results

For this research, several panel data regressions were conducted. This chapter presents the results of the regressions introduced in Chapter 2.

5.1 Basis regression

Table 3 describes the results of the first regression. It shows that the relationship between investment variables and repurchases becomes stronger after adding the control variables, except for employment. All investment variables are now significant; however, only the variable for capital expenditure is negative, while employment and R&D stay positive. This would go against expectations; however, taking in to account that this regression is subject to endogeneity problems, this regression can only be compared to the same regression in Almeida et al. (2016). They reported similar results, with the exception of employment.

Table 3

Dependent variable Capital expenditure Employment Research and

Development (R&D) Basis regression Repurchase/Assets -0.0013 (-0.43) 2.147*** (13.61) .0.0001 (1.6) N 35,494 34,177 35,477

Basis regression with control variables Repurchase/Assets -.0333*** (-10.62) 0.753*** (4.38) 0.0002** (2.34) Q 0.0014*** (34.28) 0.0619*** (26.01) -0*** (-3.52) Cashflow/Assets 0.0193*** (16.21) 0.684*** (9.34) -0.0001*** (-3.12) N 32,597 31,241 32,563

Year-quarter FE yes yes yes

Robust standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1

5.2 EPS identification strategy

Table 4 shows the results of the second regression. This regression was based on the regression of Almeida et al. (2016), but included several control variables and a broader time horizon. According to this table, an increase of repurchases with 1% of total assets leads to a decrease of capital expenditure of 0.0227%, an increase of 0.421 employment per million assets, and an increase of 0.0003% of R&D expenses. The value for the capital expenditures is consistent with expectations, the value for

(24)

employment and R&D are not. Furthermore, the results show that if the negative pre-repurchase EPS surprise increases with US $0.01, capital expenditure decreases with 0.0041% and employment increases with 0.30 employee per million assets. The effect on R&D expenses are insignificant for this variable. The value for the capital expenditure is consistent with expectations; the values for employment are not. The results for this regression are similar to those of Almeida et al. for capital expenditure, but different for employment and R&D. It should be noted that all control variables have a significant effect on the investment variables.

Table 4

Dependent variable Capital expenditure Employment R&D

Second regression

Repurchase/Assets -0.0227*** 0.421** 0.0003**

(-6.36) (2.05) (2.27)

Pre-repurchase EPS surprise 0.0052*** -0.104*** -0.000

(9.90) (-3.39) (-0.23)

Pre-repurchase EPS surprise * negative indicator

-0.0041*** 0.300*** 2.55E-05

(-6.30) (7.88) (1.22)

Dividend pay -0.0016*** -0.157*** -0.0000***

(-9.19) (-15.37) (-7.10)

Return on asset 0.0228*** 1.987*** -7.8E-05

(7.33) (10.98) (-0.78)

Debt issued 0.0158*** 0.596*** -0.0000***

(14.45) (9.31) (-2.84)

Shares issued -0.0259*** -0.0526 0.0005***

(-7.82) (-0.28) (4.45)

Quarterly stock returns 0.0048*** 0.0526* -0.0001**

(8.44) (1.83) (-3.22)

Cashflow/Assets 0.0269*** 0.621*** -0.0002***

(13.15) (5.28) (-2.80)

N 23242 22622 23216

5.3 Individual repurchases versus repurchase periods

The results of the third regression are described in Table 5. This regression is similar to the second regression, but differs by using share repurchase periods instead of individual share repurchases. The goal of this regression was to determine whether using repurchase periods produces more significant results when regressing repurchases and investments. According to Table 5, an increase in repurchases with 1% of total assets has no significant effect on capital expenditure and employment, and led to an increase of R&D expenses of 0.0011% of total assets. The increase of R&D expenses contradicts expectations. Second, Table 5 shows negative pre-repurchase EPS surprise increases with US $0.01, capital expenditure and R&D do not change significantly but it does leads to an increase in employment

(25)

of 0.52 employees per million assets. The increase of employment is in contrast with expectations. Several control variables and the negative pre-repurchase EPS surprise become insignificant.

Table 5

Third regression

Repurchase/Assets -0.0152* -0.499 0.0011***

(-1.68) (-0.64) (3.24)

Pre-repurchase EPS surprise 0.0061*** -0.230 -0.0000

(3.88) (-1.71) (-0.84)

Pre-repurchase EPS surprise * negative indicator -0.0032 0.520*** 0.0001 (-1.69) (3.17) (1.37) Dividend pay -0.0022*** -0.263*** -0.0001*** (-4.18) (-5.89) (-4.97) Return on asset 0.0388*** 3.423*** 0.0005* (5.02) (5.35) (1.87) Debt issued 0.0188*** 1.062*** -0.0002 (5.28) (3.44) (-1.50) Shares issued -0.0411*** 1.099 0.000108 (-4.67) (1.47) (0.34)

Quarterly stock returns 0.0031** 0.122 -0.0001

(2.10) (0.98) (-1.18)

Cashflow/Assets 0.030*** 1.312*** -0.0006**

(5.04) (2.63) (-2.47)

N 4211 3990 3965

5.4 Accretive repurchases

The results of the fourth regression are described in Table 6. This regression is similar to the second regression, but a dummy was added that indicates whether a repurchase is accretive. The expectation was that this indicator has a negative relationship with the investment variable, because a large share of cash holdings is needed to pay for such a repurchase, leading to a decrease in investment. It is also expected that adding an accretive repurchase indicator would decrease the effect of the negative pre-repurchase EPS surprise. According to the table, an increase of pre-repurchases with 1% of total assets leads to a decrease of capital expenditure of 0.0107%, an increase of employment of 0.88 employees per million assets, and affects R&D expenses, but not significantly. The effect on capital expenditures is consistent with the expectations, while the effect on employment is not. Second, table 5 shows that if the negative pre-repurchase EPS surprise increases with US $0.01, capital expenditure decreases with 0.0043% and employment increases with 0.292 employees per million assets. The effect on the R&D expenses are insignificant for this variable. Again, the effect on capital expenditures is consistent

(26)

with expectations, while the effect on employment is not. Third, the table shows that if a repurchase leads to an EPS increase of more than US $0.01, the capital expenditures decrease by 0.146%, employment declines 0.0554 employees per million assets, and R&D is not significantly affected. This is in accord with the expectations of the accretive indicator. When R&D removed, the coefficient negative pre-repurchase EPS surprise did not change significantly. This would indicate that there is no endogenous relation between the negative pre-repurchase EPS surprise and the accretive indicator, which is contrary to expectations.

Table 6

Fourth regression

Repurchase/Assets -0.0107** 0.880*** 0.0002

(-2.71) (3.87) (1.22)

Accretive Indicator -0.00146*** -0.0554*** 0.0000*

(-7.03) (-4.67) (1.92)

Pre-repurchase EPS surprise 0.0054*** -0.0939*** -0.000

(10.36) (-3.07) (-0.36)

Pre-repurchase EPS surprise * negative indicator -0.0043*** 0.292*** 0.0000 (-6.63) (7.66) (1.31) Dividend pay -0.0014*** -0.148*** -0.0000*** (-7.63) (-14.19) (-7.34) Return on asset 0.0239*** 2.026*** -0.0001 (7.66) (11.18) (-0.87) Debt issued 0.0155*** 0.582*** -0.0001*** (14.11) (9.07) (-2.74) Shares issued -0.0238*** 0.0226 0.0005*** (-7.19) (0.12) (4.27)

Quarterly stock returns 0.0046*** 0.0523 -0.0001***

(-8.10) (-1.63) (-3.13)

Cashflow/Assets 0.0271*** 0.626*** -0.0002***

(-13.24) (-5.33) (-2.82)

N 23242 22622 23216

Year-quarter FE Yes yes yes

(27)

6 Robustness test

In this chapter, two robustness tests are used to determine whether the regression has the same results across sector and across credit-strength levels. The expectations are the same as in chapter 3.

6.1 Repurchase effect across sectors

The summary statistics show that sectors have specific firm-characteristic and repurchase behaviors. Therefore, more insight could result from observations divided per sector. To do this, the same sectors are used as elaborated in the appendix: mining, manufacturing, TCEGS, and service, which together make 467,312 of the 542,058 observations, or roughly 86%. The second regression is used to describe the effects across sectors because this regression is the basis regression of this paper. The results for the four sectors are described in Table 7; results of the control variables were removed from the table.

Table 7

Dependent variable Capital

expenditure

Employment R&D Capital

expenditure

Employment R&D

Second regression Mining Manufacturing

Repurchase/Assets -0.246*** 0.0274 0.0000 0.0022 -0.442*** 0.0005** (-4.26) (0.12) (-0.95) (0.82) (-3.06) (2.34) Pre-repurchase EPS surprise -0.0066 -0.0229 0.0000*** 0.0038*** -0.0118 0.0000 (-1.40) (-1.27) (5.05) (10.79) (-0.62) (-0.64) Pre-repurchase EPS surprise * negative indicator 0.0141** 0.0303 -0.0000*** -0.0025*** 0.121*** 0.0001* (2.43) (1.36) (-4.59) (-5.36) (4.92) (1.78) N 1276 1236 1278 11725 11385 11677

TCEGS

Service Repurchase/Assets -0.0302 0.668 0.0000 -0.00561 -1.477*** 0.0000 (-1.10) (1.17) (.) (-1.47) (-2.57) (0.03) Pre-repurchase EPS surprise 0.0021 -0.1210** 0.0000 0.0073*** 0.0524 0.0000 (0.70) (-2.09) (.) (7.56) (-0.37) (-0.59) Pre-repurchase EPS surprise * negative indicator -0.0003 0.22*** 0.0000 -0.0061*** 0.308 0.0000 (-0.07) (2.85) (.) (-5.42) (1.82) (0.44) N 1524 1448 1521 4886 4798 4917

Year-quarter FE yes yes yes yes yes yes

The table shows that the relationship between repurchase and investment variables varies by sector. When considering the ratio of repurchases to assets, it only has a significant effect on capital expenditure in the mining sector. If the repurchase-to-asset ratio increases with 1% of total assets, capital expenditure decreases, with 0.247% of total assets in the mining sector, which is consistent

(28)

with the expectation that repurchase lead to a decrease of investments. The effect of repurchase on employment is only significant in the manufacturing and the service sectors. An increase of repurchases with 1% of total assets leads to a decrease of 0.442 employees per million assets in the manufacturing sector and 1.477 in the service sector. This is consistent with expectations. When looking at R&D expenses, the table shows that there is only a significant relation between repurchases and R&D in the manufacturing sector. However, in this sector an increase in repurchases with 1% of total assets leads to an increase of R&D expenses of 0.0005%, which goes against expectations. The effects of an increase of the negative pre-repurchase EPS surprise by US $0.01 also varies across the sectors. It significantly and negatively effects capital expenditure in manufacturing and service by 0.0025% and 0.0061%. However, it has a significant positive effect on the capital expenditure of mining of 0.0141%, which is contrary to expectations. When looking at employment, an increase of US $0.01 of negative pre-repurchase EPS surprise has a positive effect on the manufacturing and TCEGS sector of 1.21 and 2.2 employees per million assets, which goes against expectations. Finally, when looking at the relation of an increase in negative pre-repurchase EPS surprise to R&D expenses there is no significant relation, which goes against the expectations.

6.2 Repurchase effect across credit-strength levels

In addition to sectors, repurchases have a different effect on firms when their credit strength is taken into consideration (Chen and Wang, 2012). To measure the credit strength of a firm, the Altman Z-score is used. The Altman Z-Z-score is a credit-strength test that measures the likelihood of a firm going bankrupt. The formula is as follows:

𝑍 = 1.2 ∗ 𝑎 + 1.4 ∗ 𝑏 + 3.3 ∗ 𝑐 + 0.6 ∗ 𝑑 + 1.0 ∗ 𝑒 Where a= working capital / total assets

b= retained earnings / total assets

c= earnings before interest and tax / total assets d= market value of equity / total liabilities e= sales / total assets

All variables were retrieved from Compustat. After calculating the Z-scores, the firms were divided into four quartiles based on their Z-scores. Within these quartiles, the first has the lowest Z-scores and the fourth quartile has the highest. The second regression is again used as the basis regression. The results are shown in Table 8.

(29)

Table 8

Dependent variable Capital

expenditure

Employment R&D Capital

expenditure

Employment R&D

Second regression 1st_quartile ₂nd_quartile

Repurchase/Assets -0.0085 0.186 0.0003 -0.0379** 1.039** 0.0001

(-0.65) (0.27) (0.55) (-2.50) (2.56) (0.61)

Pre-repurchase EPS surprise 0.0069*** 0.0750 0.0000 0.0063*** -0.0597 0.0000

(4.16) (0.89) (-0.35) (5.11) (-1.78) (1.71)

Pre-repurchase EPS surprise * negative indicator -0.0066*** -0.0781 0.0001 -0.0049*** 0.164*** 0.0000 (-3.39) (-0.79) (1.26) (-3.25) (3.98) (-1.12) N 824 801 830 5003 4849 5002 3rd_quartile ₄th_quartile Repurchase/Assets -0.0482*** -1.077*** 0.0003 -0.0124*** -0.118 0.0001 (-8.09) (-2.82) (1.63) (-2.90) (-0.34) (-0.34)

Pre-repurchase EPS surprise 0.0048*** -0.172** 0.0000 0.0036*** 0.0471 0.0000

(6.06) (-3.37) (-0.81) (3.66) (0.59) (0.07)

Pre-repurchase EPS surprise * negative indicator

-0.0046*** 0.432*** 0.0000 -0.0022 0.238** 0.0001

(-4.53) (6.61) (0.67) (-1.71) (2.29) (0.65)

N 9796 9573 9802 7130 6911 7092

Year-quarter FE yes yes yes yes yes yes

The number of observations per quartile shows that the first quartile has a relatively low amount of share repurchases. This would indicate that firms who are relatively financially weak are less likely to repurchases shares, which is consistent with the findings of Chen and Wang (2012). The number of share repurchases increases in the second quartile, peaks in the third quartile, and decreases in the fourth quartile. Observing the effect of repurchases per asset on investments, an increase of

repurchases of 1% of the total has the following effect: a decrease in capital expenditure of 0.0379% in the second quartile, 0.0482% in the third quartile, and 0.0124% in the fourth quartile. This is consistent with the expectations. Employment increases by 1.039 employees per million assets in the second quartile and decreases by 1.21, when repurchases increase by 1% of the total assets, in the third quartile. This is only consistent with expectations for the third quartile. Repurchases per asset have no effect on R&D expenses across the quartiles. An increase of the negative pre-repurchase EPS surprise of US $0.01 has a decreasing effect on capital expenditures in the first, second, and third quartiles of 0.0066%, 0.0049% and 0.0046%, which is consistent with expectations. The increase in negative pre-repurchase EPS surprise affects employment in the second, third, and fourth quartile positively, ranging from with 0.164, 0.432, to 0.238 employees per million, respectively. Since the effect is positive, it goes against expectations. Research and development expenses are not significantly affected by the negative pre-repurchase EPS surprise in all the quartiles.

(30)

7 Conclusion, discussion, and future research

7.1 Main findings

This paper contributes to existing research by examining the effects of share repurchase on investment, using a broad time horizon of four years. The main goal of this paper was too see whether repurchases still have a significant negative effect on investments when looking at this longer, four-year, time horizon. This was done with the second regression, which formed the basis of this thesis. I found that repurchases do have an overall negative effect on capital expenditures, even when using a time horizon of four years; however, the significance of this effect varies across sectors. This research shows that when data are divided into four large sectors, only the mining sector has a negative significant relationship between share repurchases, while the rest show no significant effect. The same applies to negative pre-repurchase EPS surprises. Overall, it has significant negative effects on capital expenditures, but the effect was significantly positive for one sector, while it was significantly negative for two other sectors. Dividing the observations by credit strength showed that both repurchases and the negative pre-repurchase EPS surprise have a significant negative effect on capital expenditures. When examining the relationship between share repurchases and employment, I found an overall positive effect, though there are significant negative effects when measuring across sector. When looking at negative pre-repurchase EPS surprises, I found a significant positive relationship between the indicator and employment, both overall and across sectors. The credit-strength approach showed a negative relationship with the repurchases and positive relations between employment and the negative the pre-repurchase EPS surprise. On overall the relation between share repurchases and R&D expenses were significantly positive; this is also seen in the manufacturing sector. The relationship between employment and the negative pre-repurchase EPS surprise was negative, and this was also seen in the mining sector. However, when compared to the other investment variables, most of the R&D results were insignificant. When dividing the observations based on credit strength, none of the results from R&D were significant. As a result, I conclude that repurchase still has a significant effect on investments, in the form of capital expenditures and employment, even when considering a horizon of four years. When controlling for sector, the effects are manly negative between capital expenditure, employment, and repurchase, but do vary across sector and credit-strength group. These results are not completely consistent with the previous literature, since this suggested that share repurchases should have an overall negative effect on capital expenditures, employment, and R&D. Capital expenditures follow this line in general and on the sector level, but employment follows it only at the sector lever, but not in general. This could be explained by the fact that this research used a longer time horizon, compared to previous research, which allows firms to compensate for underinvestment

(31)

in one period with overinvestment in another. Furthermore, the flexibility with which firms can decrease and increase employment is heavily sector-dependent, since this flexibility depends on types of contracts, collective agreements, union activity, and so on.

A large variety of results were shown for R&D, when compared to the other investment variables. This can be explained by looking at the summary statistics; the number of firms with R&D expenses is limited. When using different indicators for R&D from Compustat, the results stay the same. The lack of R&D observations could influence the results. Another weak point of this research is that it only looks at the four largest sectors in the database. To gain a better insight into how repurchases effect investments across sectors, a larger set of different sectors should be used.

7.2 Other findings

This paper also aimed to examine the use of share repurchase periods instead of individual share repurchases. This has not yet been done in previous research; to examine it, a third regression was used that examined share repurchase period instead of individual share repurchases. The results of this regression show that many of the variables become insignificant when using repurchase period instead of individual repurchases. However, this is not completely sufficient to conclude that this is a better way to research repurchases. This paper limited its maximum number of successive share repurchases to 12 to avoid endogeneity problems. However, there were observations with more than 12 successive repurchases. Thus, this methodology needs to be improved so that it can use the maximum number of successions without causing endogeneity problems.

This research also looked at the effect of adding an accretive repurchase indicator to the indicator for negative pre-repurchase EPS surprises that has been used in previous research. To do this, a fourth regression was used; the results show that the accretive repurchase indicator leads to significant results, while not impacting the effects of the negative pre-repurchase EPS surprise. This goes against expectations and the conclusion of previous research that negative surprises are a predictor for accretive repurchases. This relationship can be further explored by dividing negative surprises by size, thus comparing small negative surprises with large ones. Both topics should be researched on a sector level, instead of a general level, because this study has proven that the effect of share repurchases varies by sector.

7.3 Implications of findings for future research

This paper suggests that there is a negative relationship between share repurchases and both employment and capital expenditure, over a horizon of four year. This should be taken into consideration when making repurchase decisions. However, this paper also showed that the effects vary by sector. Therefore, for future research, I suggest a broader examination of different sectors.

(32)

This research has shown that different sectors affect firms in a way that is not captured in the control variable. Thus, future research should aim to examine the effect of share repurchases on a sector-based level, instead of as an overall effect that applies to multiple sectors. Furthermore, the results of this research suggest that a different measurement for investments should be used when researching the effects on R&D. It would also be interesting to research the relationship between accretive repurchases and different negative pre-repurchase EPS surprises across sector, because previous research only describes this relationship in general. Finally, it would be interesting to do more specific research focused on comparing the difference between the use of individual repurchases and repurchase periods.

The effects of share repurchases on investment behavior